Model selection, aggregation and testing for probabilistic seismic hazard assessment

Abstract

Although the methodological framework of probabilistic seismic hazard assessment (PSHA) is well established, the selection, aggregation and testing of the underlying seismicity and ground-motion models remain a major challenge . The question to what degree the resulting hazard models can be tested has even become the topic of a controversial debate in which consensus is currently nowhere in sight. On the other hand, the last decade has witnessed considerable improvements and developments in data-driven approaches of model processing such as the use of likelihood based testing, of Bayesian inference schemes and information-theoretical model selection and ranking techniques.
An additional challenge ist the quantification of epistemic uncertainties on models. For this purpose in the context of PSHA, a logic tree framework, in which individual models are treated as alternative branches to which experts assign degree-of-belief values has become a de-facto standard. However, the assumption for the branch models to be mutually exclusive and collectively exhaustive is hard if not impossible to meet in practice, at least not in a strict sense. Alternative approaches for model aggregation, although already quite popular in other fields of science and engineering, are still waiting to be applied in the context of seismic hazard analysis.
This session is aiming to bring together people from different disciplines and backgrounds, offering different perspectives on all aspects to formally deal with models in the context of probabilistic seismic hazard analysis

Organizers

Frank Scherbaum, Universität Potsdam, Potsdam, Germany, (fs@geo.uni-potsdam.de)